Trust-sensitive computation network Nillion and decentralized open artificial intelligence infrastructure firm Ritual today announced a partnership to develop a peer-to-peer production network for AI that can process data without revealing its nature.

As AI has become even more prevalent across industries, the need to access and process sensitive information has become an equally thorny problem. This is especially an issue for industries working with personal, regulated or proprietary data that cannot be processed in the open. However, AI systems require vast computational power and large amounts of raw data to operate.

Startup Nillion builds infrastructure for peer-to-peer networks that allows decentralized information processing by mathematically sectionalizing data into nodes so that it can be computed without any single node being aware of the complete dataset. As a result, large amounts of data can be operated on “in the blind,” meaning that it cannot be leaked or revealed to third parties while distributed compute is being performed.

Ritual is a Web3 platform that uses blockchain technology to allow both traditional and blockchain-based businesses to deploy AI models on its infrastructure for executing AI tasks, known as inference. Using blockchain technology, the computation is distributed across what the company calls an AI-coprocessor on its Ritual Superchain. Ritual’s AI virtual machine facilitates the optimized execution of the model’s computation including the use of vector databases for data ingestion and graphical processing units for AI inference tasks.

One of the benefits of using Ritual as a decentralized operator for AI, the company says, is on-chain proof of interactions with models. When interacting with traditional systems, the platform can provide healthcare companies, for example, the ability to handle sensitive financial and health information for customers while providing transparency about how and where the information went through the model.

“Our partnership with Ritual is set to transform how sensitive data is processed during AI inference,” said Nillion co-founder and Chief Executive Alex Page. “Integrating our blind computation technology into Ritual’s interface will empower developers to maintain the confidentiality of both user input data and AI models, paving the way for more innovative AI applications.” 

The companies said that other opportunities for this partnership include internet of things processing, exploring price predictions that protect proprietary models and user data, and developing secure alternatives to current anonymization layers between users and generative AI chatbots and apps.

The system would also be useful for ingesting sensitive information for retrieval-augmented generation, or RAG, which uses real-time data to increase accuracy and reduce hallucinations in generative AI applications. The companies added that having secure and private access to confidential source data would open new avenues for safer and more trustworthy generative AI applications in regulated industries.

Image: geralt/Pixabay